AN EFFICIENT HYBRID LEARNING APPROACH FOR EMOTION RECOGNITION USING FACIAL EXPRESSION

Abstract

Face-based emotion identification is an important area of study in man-machine interaction research. Face accessories, uneven light, shifting settings, and other factors are some of the difficulties in the field of emotion recognition. The drawback of traditional emotion detection techniques is that feature extraction and categorization are mutually optimized. Researchers are paying more attention to deep learning (DL) techniques in an attempt to solve this problem. In classification tasks, DL approaches are becoming more and more crucial. This study addresses emotion recognition through transfer learning approaches. Nasnet Mobile Network Features with GRU-CNN (NMGC) classifier is used in this work. Finally, updating the weights is the only method available to train the newly added layers. An accuracy of 98.63% was achieved in the experiment when assigning emotions based on the CK database.